Server has 13 tools with mostly complete schemas and descriptions. Naming follows verb_noun convention (list_, get_, create_, add_, set_, remove_). Descriptions are present but vary in quality (10-150 chars). Input schemas use Zod with type definitions and descriptions. However, output schemas are not formally documented, responses are text-wrapped JSON without declared structure. Error handling is minimal; no recovery guidance or actionable error messages visible. Missing parameter descriptions for several tools (e.g., google_ads_add_keywords, google_ads_remove_keywords show no input schema in source). Composition is good, tools are single-responsibility and chainable (e.g., list_campaigns → get_campaign_structure). Security: credentials injected via environment (GOOGLE_ADS_CUSTOMER_ID, META_AD_ACCOUNT_ID), not exposed as parameters.

• Add complete input schemas for all 6 missing-schema tools (add_keywords, set_keyword_status, remove_keywords, update_bid_strategy, add_negative_keywords, set_campaign_status). Use Zod objects with .describe() on each field.
• Document output schemas formally. Define return type for each tool (e.g., 'Returns array of {id, name, status, daily_budget}' for list_campaigns). Include field types and descriptions.
• Implement actionable error messages. E.g., 'Invalid keyword match type: got "INVALID", must be one of: EXACT, PHRASE, BROAD' instead of generic failures.
• Add confirmation step for destructive operations. Require confirm_delete: true parameter for remove_keywords, or implement a dry-run mode that previews deletions.
• Expand parameter descriptions for write/update tools. E.g., for add_keywords: 'Array of keyword objects with text (string, 1-80 chars) and match_type (EXACT|PHRASE|BROAD). Max 10,000 keywords per ad group.'